Skip to main content
jg-skills
Source (opens in a new tab)

Agent Skills for professional development

Battle-tested skillsfor coding agents.

Brainstorm, plan, build, review, and ship changes with the skills professional software developers use on real projects. Workflow skills run only when you call them. Implicit skills load on their own when a task touches their stack.

Install in your project

npx skills add jesusgraterol/jg-skills --skill "*" -y

Browse the skills
  1. You: /jg-brainstorm Should we rate-limit the public API per account?
  2. Claude Code: Yes. A sliding window per account in Redis fits your middleware.
  3. You: /jg-plan
  4. Claude Code: Plan saved. Approve it to start.
  5. You: /jg-execute auto
  6. Claude Code: Built, reviewed, and pushed to main.

The skills

One skill for each step you control.

Explicit skills state exactly what they may change. Implicit skills add the stack-specific standards your agent needs on top of your repository's coding instructions.

Plan

Agree on what to build before any code changes.

  • Read-only

    jg-brainstorm

    Think an idea through before planning, with options, tradeoffs, and a recommended direction.

  • Writes planning files

    jg-plan

    Ask about the decisions that are yours, then write an implementation-ready plan for approval.

  • Read-only

    jg-challenge-plan

    Challenge a plan before approval and surface material risks, gaps, and decisions.

  • Writes planning files

    jg-breakdown

    Split a plan into ordered, reviewable milestones when it has real implementation boundaries.

Build

Turn the approved plan into verified code.

  • Changes code, can push

    jg-execute

    Build an agreed task or plan, supervised by default, or with auto to review and push each milestone.

  • Changes code

    jg-debug

    Find the cause of a bug before fixing it, then fix it with a regression test that fails first.

  • Read-only

    jg-pause

    Stop at the nearest safe point, report where things stand, and wait until you say continue.

  • Runs a picked command

    jg-command

    Turn a task into a few checked shell commands, explain each one, and run the one you pick.

Ship

Check the finished change, then publish it.

  • Read-only

    jg-review

    Review the task's changes against its plan, or the branch against a target, and give a verdict.

  • Commits and pushes

    jg-push

    Commit the current task's changes, signed when signing is set up, and push the current branch.

Context

Weigh outside input and earlier sessions.

  • Read-only

    jg-feedback

    Assess feedback independently and recommend what to adopt, reject, or adjust.

  • Read-only

    jg-session

    Find a named Claude Code or Codex session and answer questions from its latest decisions.

Implicit

Loaded by your agent when a task touches their stack.

  • Loads on its own

    jg-frontend

    Accessibility, responsive layout, themes, forms, motion, and design-system reuse for Tailwind CSS.

  • Loads on its own

    jg-react

    Components, render performance, React 19 APIs, routing, hydration, client data, and tests for React.

  • Loads on its own

    jg-pwa

    Manifests, install prompts, service worker updates, caching, and push for progressive web apps.

  • Loads on its own

    jg-seo

    Metadata, canonicals, sitemaps, robots, structured data, and social previews for public sites.

  • Loads on its own

    jg-svg

    Icon drawing, outlined logo text, file shape, brand colors, exports, and optimization for SVG.

  • Loads on its own

    jg-graphics

    Templates, fonts, exact sizes, formats, and repeatable renders for images made with Playwright.

  • Loads on its own

    jg-nestjs

    Modules, schema validation, configuration, logging, lifecycle, and tests for NestJS.

  • Loads on its own

    jg-postgres

    Naming, constraints, queries, transactions, pools, timeouts, and migrations for PostgreSQL.

  • Loads on its own

    jg-redis

    Key names, expiry, scanning, round trips, data structures, and locks for Redis.

  • Loads on its own

    jg-node

    Cross-platform paths, safe filesystem access, processes, and CLIs for Node.js scripts.

  • Loads on its own

    jg-packages

    Package selection, local API documentation, and consumer integration for JG Packages.

  • Loads on its own

    jg-npm-package

    Exports, dependency ranges, tarball tests, versioning, and trusted publishing for npm packages.

  • Loads on its own

    jg-monorepo

    Task graphs, cache inputs, filters, and workspace dependencies for Turborepo and pnpm.

  • Loads on its own

    jg-bash

    Shell choice, script structure, quoting, safety, and verification for bash and sh on Linux.

  • Loads on its own

    jg-docker

    Compose services, healthchecks, secrets, ports, image tags, and Dockerfile structure for Docker.

  • Loads on its own

    jg-ci

    A single check workflow, permissions, timeouts, caching, and action pinning for GitHub Actions.

  • Loads on its own

    jg-zod

    Strictness, metadata, error handling, formats, and composition for Zod 4 schemas.

  • Loads on its own

    jg-testing

    Deterministic waits, timeouts, retries, and repeat runs for Vitest and Playwright.

How it works

From a conversation to a pushed commit.

How jg-execute runs
  1. 01

    Plan

    Agree on what to build before any code changes.

    1. /jg-brainstorm

      Agree on a direction.

    2. /jg-plan

      Save it as a plan you approve.

    3. /jg-challenge-planOptional

      Find gaps before you approve.

    4. /jg-breakdownOptional

      Split a large plan into milestones.

  2. 02

    Build

    Turn the approved plan into verified code.

    1. /jg-execute

      Build and check one milestone, then stop.

    2. /jg-debugOptional

      Find a bug's cause, then fix it test-first.

    3. /jg-pauseOptional

      Stop at a safe point, then pick up later.

    4. /jg-commandOptional

      Pick a checked command and run it.

  3. 03

    Ship

    Check the finished change, then publish it.

    1. /jg-review

      Check the change before it ships.

    2. /jg-push

      Commit and push the task's files.

    You run these when you are ready.

Install

Add them to your project.

Run the commands from your repository's root. The skills live in the project, so you can commit them and share the same workflow with your team.

Needs Node.js. Explicit-only and implicit loading are verified with Claude Code and Codex.

Update refreshes only the skills you already have. To add skills released since you installed, run the install command again.

Install

npx skills add jesusgraterol/jg-skills --skill "*" -y

Update

npx skills update -p jg-brainstorm jg-plan jg-challenge-plan jg-breakdown jg-execute jg-debug jg-pause jg-command jg-review jg-push jg-feedback jg-session jg-frontend jg-react jg-pwa jg-seo jg-svg jg-graphics jg-nestjs jg-postgres jg-redis jg-node jg-packages jg-npm-package jg-monorepo jg-bash jg-docker jg-ci jg-zod jg-testing

Remove

npx skills remove jg-brainstorm jg-plan jg-challenge-plan jg-breakdown jg-execute jg-debug jg-pause jg-command jg-review jg-push jg-feedback jg-session jg-frontend jg-react jg-pwa jg-seo jg-svg jg-graphics jg-nestjs jg-postgres jg-redis jg-node jg-packages jg-npm-package jg-monorepo jg-bash jg-docker jg-ci jg-zod jg-testing

Agentic Coding

Give your agent standards to follow.

The skills run the workflow. Your repository's AGENTS.md sets the standards each one applies, and Agentic Coding has a starting point either way.

  1. AGENTS.mdSets the standards
  2. jg-skillsRuns the workflow

jg-packages

Give your agent reusable building blocks.

Use reusable helpers alongside your project's own code. The jg-packages skill helps your agent choose a suitable package and verify its installed API.

Browse jg-packages (opens in a new tab) Use the jg-packages skill

  1. Coding instructions

    Set your project's standards.

  2. jg-skills

    Guide the work.

  3. jg-packages

    Provide reusable code.

Credits

Ideas we built on.

A few ideas in these skills are adapted from other open-source collections and rewritten in our own words.